Janus-type dual emission of a Cyclometalated Iron(III) complex

Photoactive compounds are essential for photocatalytic and luminescent applications, such as photoredox catalysis or light emitting diodes. However, the substitution of noble metals, which are almost exclusively used, by base metals remains a major challenge on the way to a more sustainable world.1 Iron is a dream candidate for this ambitious aim.2 But compared to noble metal complexes that show long-lived metal-to-ligand charge-transfer (MLCT) states, realization of emissive and photoactive iron complexes is demanding, due to the fast deactivation of charge transfer states into non-emissive inactive states. No MLCT emission has been observed for monometallic iron complexes before. Consequently, dual emission could also not yet be realized with iron complexes, as it is a very rare property even of noble metal compounds. Here we report the Fe III complex [Fe(ImP) 2 ][PF 6 ] (HImP = 1,1’-(1,3-phenylene)bis(3-methyl-1-imidazol-2-ylidene)), showing Janus-type dual emission by combining LMCT (ligand-to-metal charge transfer) with MLCT luminescence. The respective excited states are characterized by a record lifetime of τ MLCT = 4.2 ns, and a moderate τ LMCT = 0.2 ns. Only two emissive Fe III compounds are known so far and they show LMCT luminescence only.3,4 The unique properties of the presented complex are caused by the specic ligand design combining four N-heterocyclic carbenes with two cyclometalating groups, using the σ-donor strength of six carbon atoms and the acceptor capabilities of the central phenyl rings. Spectroscopically, doublet manifolds could be identied in the deactivation process, while (TD)DFT analysis revealed the presence of quartets as well. With three key advancements of realizing the rst iron complex showing dual luminescence, a MLCT luminescence and a world record MLCT lifetime, the results constitute a basis for future application of iron complexes as white light emitters and new photocatalytic reactions making use of the Janus-type properties

in Fe III complexes due to the di cult accessibility of iron(IV) in typical coordination complexes.
Cyclometalating phenyl-containing ligands offer both strong σ-donor but also π-donor properties which stabilize high oxidation states. Fe II complexes with such ligand types were extensively theoretically by Jakubikova and Dixon. [18][19][20][21] The quantum chemical predictions were recently supported by an experimental study from our groups. 22 The Fe II complex, derived from [Fe(tpy) 2 ] 2+ (tpy = terpyridin) by exchange of one tpy through a deprotonated phenylbipyridine, showed an extension of the MLCT lifetime by a factor of 5 and a decrease of the MC lifetime. Here we explore the application of cyclometalating ligands in iron(III) complexes to extend CT lifetimes, which nally results in a two-color luminescence that has never been observed for iron complexes before.
In the crystal structure (Fig. 1b), the C NHC -Fe-C NHC bite angle of 155° is smaller than in the analog terpyridine complex 8 and leads to a more distorted octahedral geometry. The ligand to ligand dihedral angle is 86°. One counterion is present in the unit cell, in agreement with a Fe III center. The doublet shown in the Mößbauer spectrum of Fig. 1c is characterized by an isomer shift of -0.12 mm s -1 and a quadrupole splitting of 1.59 mm s -1 , indicating a d 5 low-spin complex. Temperature dependent magnetic susceptibility measurements (Extended Data Fig. 3) show the typical behavior of a low-spin Fe III compound, i.e. the χ M T values are higher than the spin-only value (0.375 cm 3 mol -1 K vs. 0.49 -0.64 cm 3 mol -1 K in 1) and exhibit the expected deviation from the Curie law due to non-quenched orbital momentum of the 2 T 2 ground state.
Beyond enforcing a low-spin con guration, the tremendous effect of cyclometalation and the NHC ligands on the electronic structure is revealed by cyclic voltammetry (Fig. 2a). A reversible Fe II/III redox wave is found at very low potential (-1.16 V vs FcH 0/+ ). Compared to the value of 0.31 V for the analogous iron complex with two C^N^C ligands (2,6-bis(3-methyl-imidazole-1-ylidine)-pyridine), a cathodic shift of 1.47 V is observed. 26 This corresponds to the behavior of the [Fe(tpy) 2 ] 2+ / [Fe(C^N^N) (tpy)] + pair with a cathodic shift of 0.83 V. 22 A quasi-reversible wave at 0.08 V is assigned to the Fe III/IV couple, while the irreversible wave at E p = 1.23 V is attributed to ligand oxidation (Extended Data Fig. 4a).
Both MLCT (λ ε,max = 351 nm, ε= 6,000 M -1 cm -1 ) and LMCT (λ ε,max = 585 nm, ε= 540 M -1 cm -1 ) transitions are observed in the absorption spectrum of 1 (Fig. 2b). Spectra of the electrochemically generated oneelectron reduced and oxidized species 1and 1 + are shown in Extended Data Fig. 4b. According to DFT calculations, these are metal-centered redox processes, showing the mainly Fe II and Fe IV character of 1 -and 1 + , respectively (Extended Data Table 1). TDDFT calculations with optimally-tuned range-separated functionals suited for the description of charge-transfer states 27,28 further reveal the nature of the underlying transitions ( Fig. 2b and 2c, Extended data Fig. 5a and 5b). The LMCT absorption is caused by a transition from the ligand π-orbital involving both imidazole and phenyl donors to a singly occupied metal-centered d-acceptor orbital of the t 2g set. In the MLCT band, transitions originate from singly and doubly occupied Fe III d-orbitals to the π*-orbitals of the cyclometalating phenyl ligand. Some transitions in the MLCT region from 300-450 nm are of mixed LMCT/LC character to a certain extent.
Excitation of 1 into the low energy LMCT absorption band at 585 nm results in a broad emission mirroring the LMCT absorption band (Fig. 3a). On the other hand, excitation into the high energy MLCT absorption at 350 nm results in dual room temperature luminescence (Fig. 3a). The high energy emission at λ max = 450 nm and a broad band emission at λ max = 675 nm re ect the MLCT and LMCT absorption bands, respectively. From the rather small Stokes shifts, doublet 2 MLCT and 2 LMCT are deduced as emissive states. The excitation spectra shown in Extended Data Fig. 6a recorded with λ em = 420 nm and 675 nm largely follow the absorption spectrum, supporting that the observed dual photoluminescence originates from complex 1. A two-dimensional excitation -emission plot is shown in the Extended Data Fig. 6b. In agreement with the two-color absorption composed of MLCT and LMCT bands, the high energy MLCT emission vanishes with decreasing excitation energy. With decreasing temperature, the emission bands increase in intensity (Extended Data Fig. 7). Dual emission is generally a very rare observation, 29 and it is the rst time that such a behavior is reported for an iron compound at all. Since luminescent complexes of earth-abundant elements are also still very rare, 30 the observation of a Janus-type dual MLCT and LMCT emission is an encouraging development for future substitution of noble by base metals for photochemical applications.
A detailed view on the excited-state landscape of complex 1 is extracted by ultrafast spectroscopy. The transient absorption (TA) spectra after excitation at 330 nm and the decay associated amplitude spectra signatures of neither the Fe II nor Fe IV species. It might be caused by an electronic relaxation from higher lying MLCT states. TA measurements applying pump pulses into the LMCT absorption band at 600 nm, shown in extended data gure 8, reveal the same ESA bands observed at 330 nm excitation but a single exponential signal decay. The time constant is 0.24 ns proving the assignment to an 2 LMCT state, which is directly optically excited here.
Streak camera measurements (SCM) with excitation at 330 nm (time resolution 50 ps) reveal a luminescence in the spectral region above 640 nm, re ecting the 2 LMCT emission. An additional signal in the spectral range 390 to 600 nm, which extends over a few nanoseconds, corresponds to the luminescence above attributed to the 2 MLCT states. A time constant of 0.22 ns is found for the 2 LMCT emission, which is in excellent agreement with the TA results. For the 2 MLCT luminescence, two time constants of 2.1 ns and 5.2 ns are obtained. The spectrally integrated signal of the 2 MLCT emission and the corresponding t are shown in Fig. 3c. Time correlated single photon counting (TCSPC) experiments verify this result (extended data gure 9), since two decay components with time constants of around 2.1 ns and 9 ns for the 2 MLCT emission are obtained. Amplitude spectra for the two decay components, shown in Fig. 3d, were extracted from the SCM data by a global t. Since the two spectra are basically identical, the two emission components have to originate from the same state. This might indicate that the decay is non-exponential, and the double exponential t can only approximate the real conditions. It is thus valid to deduce an average MLCT lifetime of 4.2 ns. Speculatively, this effect can be attributed to the exibility of the ligands which cause a distribution of slightly different conformers. The 2 MLCT decay is likely to be sensitive to structure variations in the ensemble of complexes probed.
The spectroscopic results, combined with ground state DFT calculations on the doublet, quartet and sextet state in their respective optimized geometry are summarized in the schematic excited state landscape given in Fig. 4a. Optical excitation in the near UV addresses hot 2 MLCT states. Electronic relaxation exhibits a branching shortly after the excitation. The population majority is transferred within 0.5 ps to the lowest 2 LMCT state. This state is emissive but decays to the ground state mostly by internal conversion resulting in a lifetime of 0.2 ns. The weak absorption band at 600 nm, associated with the 2 LMCT state, shows that the corresponding transition dipole moment is small. In combination with the limited lifetime, a low quantum yield of less than 1 % results, as estimated by the intensity of the solvent Raman bands (Extended Data Fig. 6). A minor fraction of the excited population stays in the MLCT manifold and relaxes to the lowest 2 MLCT state. The 6 ps component in the TA measurements can be a signature of this relaxation. The 2 MLCT state exhibits a lifetime of 4.2 ns and relaxes non-radiatively as well as radiatively back to the ground state, resulting in the weak but signi cant MLCT emission in the blue spectral region. The ground state DFT calculations show that a sextet MC state can be most likely excluded from the deactivation pathway since it is strongly destabilized by the tremendous donor properties of the ligands. However, the energy of the 4 MC state in this approximation is lower than for the 2 MLCT and slightly higher than for the 2 LMCT. Therefore, 4 MC states cannot be excluded from deactivation pathway, although no spectroscopic signature could be detected. But they could act as a quencher for both MLCT and LMCT states and thus explain the observed quantum yield.
The scenario summarized in Fig. 4a is further supported by quantum-chemical TDDFT calculations of potential energy curves, Huang-Rhys factors, and nonadiabatic couplings performed in a constrained way in D 2d symmetry (cf. Extended Data Fig. 5c). This approach indicates that the description of the dynamics by a few-state scheme is indeed a valuable and easy interpretable simpli cation, which is already indicated in the TCSPC and SCM discussion. Nevertheless, the qualitative picture derived from the potential energy curves along the symmetric a 1 mode having the strongest Huang-Rhys factor in the given energy range ( Quartet states qualify for participation in this relaxation channel and support the claim for being responsible for the low quantum yield. The second pathway leading to an emissive 2 MLCT state would require a transient structural and electronic stabilization. A possible candidate for such a state has been tentatively assigned in Fig. 4b. Here, already rather small energetic corrections to the potential curves beyond TDDFT and further low symmetry modes, which would in particular stabilize LMCT states, could increase the barrier due to the crossing curves such as to provide a transient trapping of population.
Future work has to eliminate the bottlenecks of the limited accuracy of the TDDFT method and identi cation of the relevant vibrational modes to allow for a quantitative theoretical analysis.
In summary, coordination of phenylene-bis-imidazolylidene ligands to a Fe III center yields an air-and water stable doubly cyclometalated iron complex that shows an unprecedented Janus-type two-color luminescence from both MLCT and LMCT states, respectively. Such a unique behavior originates from the combination of strong NHC and cyclometalating s-donors to cause LMCT luminescence. The additional acceptor-capabilities of the cyclometalating ligand enable the MLCT emission. This is the rst time that dual luminescence is reported for an iron complex, but also MLCT emission has not been observed for mono-nuclear iron compounds before. 17 While the LMCT state already has a considerable lifetime of 0.2 ns, the MLCT state shows an even longer lifetime of 4.2 ns, which is the longest measured CT lifetime for iron complexes so far. With these results, the foundations for the usage of earth-abundant iron and its complexes in numerous photochemical and photophysical applications are set. Based on the presented data, iron-based white light emitters and multifunctional photoredox catalysts become accessible.  further precipitation of a pale solid was observed. The suspension was ltered through a cotton ball and afterwards through a porous glass frit. The respective lter cakes were washed with acetonitrile until the ltrate turned colorless. The solvent of the ltrate was evaporated using a rotary evaporator. The blue solid was dissolved in dichloromethane and ltered over a silica column. The column was washed thoroughly with dichloromethane. The blue band was eluted with acetonitrile. The solvent of the blue fraction was evaporated. The solid was dissolved in methanol and KPF 6 (2 eq., 736 mg, 4 mmol) were added. The desired compound 1 was precipitated by the addition of water and ltered off. It was redissolved in methanol and treated again with KPF 6 (2 eq) and precipitated again with water to ensure full replacement of the counterion. The suspension was ltered, and the blue solid was dried thoroughly under vacuum. It was then dissolved in a minimal amount of dichloromethane and pentane was allowed X-ray diffraction analysis.
The single crystal data were recorded using a Bruker SMART CCD area-detector diffractometer equipped with a graphite monochromator. The measurements were carried out using Mo Kα radiation (λ = 1.54178 Å) at T = 200(2) K, since at lower temperatures a phase transition occurred, which caused a vaguer diffraction pattern. Structure solution was carried out by direct methods 4 and structure re nement was conducted using full-matrix least squares re nement based on F². 4 All non-H-atoms were re ned anisotropically and the hydrogen atom positions were derived from geometrical reasons -except hydrogens of methyl groups. They were located from Fourier map using HFIX 137 by SHELX. 4 All hydrogen atoms were re ned at idealized positions riding on the carbon atoms with isotropic displacement parameters U iso (H) = 1.2U eq (C) resp. 1.5U eq (-CH 3 ) and C-H bond lengths of 0.93-0.96 Å. All CH 3 hydrogen atoms were allowed to rotate but not to tip. One dichloromethane solvent molecule could not be modelled during re nement and was treated using SQUEEZE from the platon software package. [5][6][7] (C 28 H 26 N 8 Fe)(PF 6  Mößbauer spectroscopy. Mößbauer spectra were recorded with a 57 Co source in a Rh matrix using an alternating constant acceleration Wissel Mössbauer spectrometer operated in the transmission mode and equipped with a Janis closed-cycle helium cryostat. Isomer shifts are given relative to iron metal at ambient temperature. Simulation of the experimental data was performed with the M t program using Lorentzian line doublets: E. Bill, Max-Planck Institute for Chemical Energy Conversion, Mülheim/Ruhr, Germany.
Magnetic susceptibility measurements. Temperature-dependent magnetic susceptibility measurements were carried out with a Quantum-Design MPMS3 SQUID magnetometer in the range from 300 to 2.0 K at a magnetic eld of 0.5 T. The powdered sample was contained in a polycarbonate capsule and xed in a non-magnetic sample holder. Each raw data le for the measured magnetic moment was corrected for the diamagnetic contribution of the sample holder and the capsule. The molar susceptibility data were corrected for the diamagnetic contribution.
Cyclic voltammetry and square wave voltammetry. Absorption spectroscopy (steady state).
Acetonitrile of spectroscopic grade (Spectronorm VWR Acetonitrile) was used as solvent for steady-state absorption spectroscopy.
Steady state absorption spectra were recorded using solutions with concentrations of about 10 -5 M in quartz-cuvettes (pathlength 10 mm) by a Cary 50 spectrometer.
Room temperature emission spectroscopy.
For steady-state emission spectroscopy acetonitrile of spectroscopic grade was used as solvent.
Steady-state emission spectra were recorded in 10 mm quartz cuvettes on a Jasco FP8300 or a Horiba Scienti c FluoroMax-4 spectrometer. The solutions for the measurements under argon were degassed via the freeze-pump-thaw technique.
Variable temperature emission spectroscopy.
Variable-Temperature Emission spectra were recorded on a Varian Cary Eclipse spectrometer. For low temperature photoluminescence measurements, a solution of the complex in butyronitrile (re uxed over Femtosecond transient absorption spectra were recorded using excitation wavelengths in three different optical regions and thereby somewhat different pump-probe setups. In all cases they are based on regenerative Ti:sapphire laser systems operating at a frequency of 1 kHz and at a center wavelength of 775 nm (CPA 2001, Clark MXR, Inc.) respectively 800 nm (Spit re Pro, Spectra-Physics). For probing, a white light continuum generated by focusing a small fraction of the Ti:sapphire output into a CaF 2 crystal was used. Pump and probe beam were focused onto the sample to overlapping spots with diameters in the range of 200 to 400 µm for the pump and of 100 µm for the probe. The polarizations of the pump and probe pulses were set to magic angle with respect to each other. After the sample, the probe was dispersed by a prism and transient absorption changes were spectrally resolved recorded by an array detector.
For pumping the sample with an excitation wavelength of 400 nm the output of the Ti:sapphire system (Spit re Pro) was frequency doubled by a BBO crystal. The resulting time resolution was about 150 fs.
To obtain ultrashort excitation pulses in the visible with a center wavelength of 600 nm a non-collinear optical parametric ampli er (NOPA) pumped by the Ti:sapphire system (CPA 2001) was applied. The dispersion of the NOPA pulses was minimized by a compressor based on fused silica prisms resulting in an overall time resolution of better than 100 fs.
For excitation in the UV, i. e., at a center wavelength of 330 nm, the NOPA was tuned to 660 nm and its output was frequency doubled by 100 µm thick BBO crystal cut for type I phase matching.
For all measurements, the iron complex was dissolved in acetonitrile under argon and the sample solution was lled into a fused silica cuvette with a thickness of 1 mm.
The obtained data was tted using a global t. In the global t, the multi-exponential model function , convoluted with the temporal response of the pump-probe setup, is tted to the complete set of time dependent transient absorption spectra. In the present case three exponential decay components were necessary to reproduce the data with satisfying accuracy, i. e. N = 3 Streak Camera Measurements.
In addition to the TCSPC measurements, the time resolved luminescence was also investigated applying a streak camera (Streakscope C10627, Hamamatsu Photonics). The samples were prepared and measured under argon in 1 cm cuvettes. For excitation at 330 nm, a NOPA pumped by a Ti:sapphire laser system (CPA 2001, Clark MXR, Inc.) was set to a center wavelength of 660 nm and its output pulses were frequency doubled by a BBO crystal. To ensure that only radiation at 330 nm reaches the sample, a fused silica prism was applied to separate the UV pulses from the fundamental.
The luminescence lifetimes were determined by tting a monoexponential decay to the data in the spectral region 640 to 840 nm and a double exponential decay to the data of the region 390 to 600 nm.
Quantum-chemical calculations were performed at D 2d symmetry with DFT and linear response TDDFT using the optimally-tuned long-range separation functional LC-BLYP together with combined basis set: def2TZVP (Fe) and 6-311G(d,p) (all other atoms). Tuning of the functional was done with the so-called ΔSCF method [13][14][15] , the details can be found in the work of Bokarev et al. 16 The following parameters were obtained for the present complex: α=0.0 (percentage of exact exchange in short-range) and 0.15 bohr -1 (long-range separation parameter). Solvent effects (acetonitrile) were taken into account within the polarized continuum model (PCM) approach. 17 Calculations were done with G16 18 and the Q-Chem 5. 3 19 packages. Excited state analysis was performed using the TheoDORE package. 20 Analysis of Huang-Rhys factors, tuning of functional, and generation of geometries along normal modes were done with inhouse codes.

Data Availability
The data generated and analyzed during the current study are available from the corresponding author on reasonable request. X-ray source data for g. 1b are deposited under CCDC identi er CCDC 2002774 and are available in the supporting information.

Code Availability
The code used for analysis of Huang-Rhys factors, tuning of functional, and generation of geometries along normal modes is available from O.K. upon reasonable request.  centered π-π* (LC), metal-centered (MC), ligand-to-metal charge transfer (LMCT) and metal-to-ligand charge transfer (MLCT) states. c, Molecular orbital scheme showing the highest occupied orbitals (t2g orbitals highlighted red, ligand-based orbitals blue) and the lowest unoccupied orbitals (π* orbitals of the phenyl moiety highlighted green). The transition densities of the dominant LMCT (above) and MLCT (below) transitions are also depicted here (hole (blue) and electron (red)).